Discrete subsystem selection using linear physical programming

Abstract

The design of complex systems may involve The design of complex systems may involve the the selection of several subsystem designs. In this paper, we investigate the problem of selecting discrete concepts from multiple, coupled subsystems. This problem is one where_ both subsystem level (local) measures of merit and system level (global) measures of merit are present, We develop an approach to obtain the sets of subsystem design concepts that will satisfy the system objectives. Graph Theory is used to represent the coupled selection problem where the nodes of the graph are the subsystem design choices and the arcs connecting the nodes indicate the relationships between the subsystems. Discrete optimization techniques from graph theory and linear physical programming are combined to form a powerful algorithm to solve this problem. The approach presented in this paper can be used by a designer to decrease the number of subsystem combinations that represent ,successful system design alternatives. Such a tool can be most useful at the conceptual design stage of the design process where the number of design alternatives is potentially very high and the need for identifying successful subsystem design combinations arises. Once the promising subsystem designs are obtained at the conceptual design stage, focus can be restricted on these chosen design alternatives for further testing and refinement at a later embodiment design stage. selection of several subsystem designs. It is of significant importance that the subsystem designs selected are compatible with one another in order for the overall system to function efficiently. Considerable work has gone into selection of concepts at the various stages of the design process. Ulrich and Eppinger’ use the Pugh selection method in comparing designs at the concept screening stage with the help of decision matrices. At the concept scoring stage, the designer weights the various concepts based on the customer preferences. The concept scores are determined by the weighted sum of the ratings. Hazelrigg’ also details the use of utility theory and von Neumann-Morgenstern lotteries to make a choice amon 5 certain given number of design options. Yang and Sen use a hierarchical evaluation process for multi-attribute design selection with uncertainty. They use hierarchical factor structure for evaluation and quantification of a qualitative attribute. But most methods while handling multi-attribute selection problems do not address the selection of a design from multiple coupled subsystems that have conflicting objectives. Also, many methods identify a single best concept. On the other hand, we feel that at the conceptual design stage, identifying a handful of good designs could prove to be a more promising product development exercise. Therefore, we present an approach to identify a number of good subsystem design combinations based on local and global measures of merit.